Table 2:

Relationship between HDHP Offer and Total Family Health Care Spending

	(1)	(2)	(3)	(4)
	Total family health care spending	Health care spending for member with SUD	Health care spending for other members	Spending on SUD services
Average percentage point change in probability of any spending (95% CI) attributable to HDHP offer
Post X Treatment	−0.014 (−0.031, 0.004)	−0.014 (−0.031, 0.003)	−0.015* (−0.031, 0.001)	−0.024*** (−0.038, −0.010)
Observations	317,353	317,353	317,353	317,353
R-squared	0.246	0.111	0.218	0.143
Pre-HDHP offer mean	0.990	0.949	0.902	0.392
Average change in conditional spending (95% CI) attributable to HDHP offer
Post X Treatment	−1,545.78*** (−2,270.86, −820.70)	−1,184.79*** (−1,844.79, −525.20)	−359.00*** (−602.32, −115.68)	72.52 (−153.35, 298.38)
Observations	313,859	298,740	287,879	108,431
R-squared	0.366	0.259	0.262	0.167
Pre-HDHP offer mean	20,226	13,826	7,655	3,724

Notes: The relationship between the family HDHP offer and the probability of any spending and spending, conditional on any spending are shown. Post X Treatment coefficients were calculated a difference-in-differences framework comparing families at treatment (HDHP offer) vs. comparison employers (no HDHP offer) with employer and year fixed effects. All models were adjusted for the documented age, sex, and race/ethnicity of the policyholder, family income and median Census block education levels, the number of family members, and the presence of chronic conditions among any family member derived from the Chronic Conditions Warehouse (CCW) methodology. Probability of any spending were modeled as linear probability models. Spending, conditional on any spending, was log-transformed and coefficient estimates were exponentiated to obtain relative spending coefficients. Pre-HDHP offer means were calculated among families at treatment employers prior to the HDHP offer. All measures were constructed from the OptumLabs Data Warehouse for the years 2007 to 2017.